Channel estimation and data detection for downlink multicarrier CDMA mobile systems.

Mahmoud Mir Ali Hussain Al-Quzwini
Dr. Sarcote Namik Abdulah

Abstract: Performance and flexibility of mobile radio systems essentially depend on the choice of the multiple access scheme. The novel Multicarrier Code Division Multiple Access (MC-CDMA) scheme which is the topic of this thesis is a promising alternative to conventional FDMA, TDMA, and CDMA schemes. With MC-CDMA, each data modulated chip of the user specific spreading code is mapped onto a separate subcarrier, so that the chips of the spreading code and, hence, the data symbol is transmitted in parallel on multiple subcarriers which introduces frequency diversity. The application of OFDM enables one to avoid ISI and ICI in the detection process, resulting in simple Single User Detection (SD) techniques. In MC-CDMA mobile radio systems, SD can be realized by equalizing each subcarrier individually, yielding only one complex-valued multiplication per subcarrier. Due to the M-, Q-, and M&Q Modification of the MC-CDMA scheme, flexible MC-CDMA systems can be designed to support different applications with different data rates and different number of active users. Coherent detection of MC-CDMA signals dictates that a per-subband estimation of the channel frequency response is generated for each OFDM symbol. In this thesis, a spectrum efficient decision directed adaptive MC-CDMA mobile receiver for the down link is developed and simulated. It consists of a Minimum Mean Square Error (MMSE) equalizer and an Improved Recursive Least Square (ImRLS) adaptive channel estimator. The MMSE equalizer is shown to be able to maintain user orthognality with low average BER. The ImRLS is developed from the well known RLS adaptive channel estimator by time limiting the estimated channel frequency response using IFFT/FFT operations. Performance evaluation in COST 259 channel models, Typical Urban (TU) and Hilly Terrain (HT), show that the ImRLS adaptive estimators outperforms the RLS adaptive channel estimator in terms of average BER, estimation average mean square error , the ability to track the channel variations at high mobile velocity, and the ability to handle heavily loaded systems. While RLS estimator was only able to respond to the channel variations at a mobile velocity =3 km/h, the ImRLS estimator is able to effectively handle mobile velocities = 90km/h. The ImRLS with frequency domain interpolation ImRLS-P is also proposed to reduce the computational complexity of the ImRLS adaptive channel estimator. The strategy adopted by the ImRLS-P is to estimate the channel at only certain subcarriers, which are much less than the total subcarriers, then the rest part of the channel frequency response can be interpolated out using FFT interpolation. Performance evaluation shows that the ImRLS-P estimator can achieve a comparable performance to that of the ImRLS with only 50% of the total subcarriers employed for channel estimation. The Pilot Symbol Aided (PSA) channel estimation technique with rectangularly spaced pilot symbols is described and simulated. Simulation results show that for mobile velocities = 90 km/h, only the PSA channel estimator can achieve adequate estimation accuracy but at the expense of sacrificing 12.5% of the available bandwidth and the total number of active users. Simulations are carried out using the MATLAB®2007a software package.